CN101512940A - Method and equipment for estimating effective noise substrate of memory - Google Patents

Abstract

A method for noise rise estimation in a wireless communications system is presented, which comprises measuring of samples of at least received total wideband power. From the measured samples of at least received total wideband power, a probability distribution (800) for a first power quantity is estimated. This first power quantity can be the received total wideband power itself. The probability distribution (800) for the first power quantity is used for computing a conditional probability distribution (803) of a noise floor measure. This computing is performed recursively. A value of a noise rise measure is finally calculated based on the conditional probability distribution for the noise floor measure. A node of a wireless communications system having the above functionality is also presented. Typically, the node is a RNC.

Description

The method and apparatus that is used for the memory efficient noise floor estimation

Technical field

The present invention relates generally to the relevant amount of power that is used for estimating cellular communication system, particularly, is used for the method and apparatus of estimating noise substrate.

Background technology

Wideband Code Division Multiple Access (WCDMA) (WCDMA) telecommunication system has many attractive features, can be used in the exploitation of telecommunications service in the future.WCDMA for example and similarly in the system concrete technological challenge be the uplink channel that strengthens to the scheduling in the time interval, wherein disturbed condition be favourable and wherein the up link in the sub-district of being discussed exist enough capacity to be used for supporting the uplink channel that strengthens.As everyone knows, the user of the existence of sub-district is the up link contribution interference level of WCDMA system.And the terminal in adjacent sub-district is also contributed identical interference level.This is because when using CDMA technology, all users and the common channel of sub-district send in same frequency range.The load of sub-district directly relates to the interference level of same sub-district.Therefore the permission controlled function of RNC is concentrated in WCDMA, because overload causes the service quality and the unsettled sub-district of difference, these behaviors need be avoided.

The field is estimated in the load that the present invention relates in CDMA cellular telephone system.Several provided for radio resources management (RRM) algorithm such as scheduling and permission control, depends on the accurate estimation of uplink load.

The permission control algolithm needs the available resources and the traffic carrying capacity of being asked by the user of each sub-district of balance or RBS.This means, comprise available HW resource to the important input of permission control algolithm, and about the information of number of users instantaneous in each sub-district and theys' ongoing traffic carrying capacity.

In order to keep sub-district stability and increase capacity, the enhanced uplink dispatching algorithm is used for keeping load to be lower than certain level fast.This is because most of uplink users channels in WCDMA, are subjected to power control at least.The purpose of this power control is to keep the power level of reception of each channel at certain wanted to interfering signal ratio (SIR), so that make it possible to satisfy specific service request.This SIR level makes that normally the received power in radio base station (RBS) is than the low several dB of interference level.The spread spectrum of separating in so-called RAKE receiver strengthens each channel then to such signal level, and wherein the bit of Fa Songing for example can be for further processing by channel decoder that is arranged in the signal processing chain back and speech coder and decoder device.

Because it is its specific preferred sir value that RBS attempts keeping each channel, may occur, additional user, or existing user's bursty data traffic carrying capacity cause interference level, the instantaneous thus SIR that reduces for other user.The response of RBS is that other all user of order power increases, and this causes to disturb further sometimes increases.Usually, this process keeps stable being lower than under certain load level.Under the emergent situation of high capacity channel, it is very big that the rising of interference becomes, thereby increase instable risk, and so-called power emerges in large numbers.Therefore be necessary to dispatch the high power capacity uplink channel,, like this, can guarantee to avoid unsteadiness as the enhanced uplink in WCDMA (E-UL) channel.In order to accomplish this point, must estimate the instantaneous load in RBS.This enables to assess the capacity surplus from point of instability.

Useful especially tolerance is up link (and down link) cell load that the rising (or noise rising) according to heat is measured.The rising of heat (ROT) is defined in the merchant between instantaneous broadband power and the thermal noise floor level.All noise rise measure have common ground: they depend on the accurate estimation of background noise.According to the quantity of power of the height fluctuation of prior art or noise floor determine typically be associated with sizable uncertainty, these uncertain or even with whole available identical magnitudes of capacity surplus.Therefore, under the situation of not improving the load estimation that is connected to it, the very difficult really uplink channel function of implementing enhancing.

At this moment, can propose, the no less important parameter that needs load to estimate for its control is the coverage of sub-district.This coverage be usually directed to for operate as normal need be with the special services of specific SIR operation.The uplink cell border is defined by the terminal with the peak power output operation then.Maximum receive channel power in RBS by terminal maximum power and be defined to the path loss of digital receiver.Because path loss is the direct function of the distance between terminal and RBS, draws the ultimate range from RBS.This distance definition coverage of on all directions, getting from RBS.

Draw now, any increase of interference level causes the SIR that reduces, and it can not be compensated by the terminal power that increases.Therefore, need reduce path loss, to keep service.This means that terminal need move on to more close RBS, that is, reduce the coverage of sub-district.

From above discussion, can know and see, in order to keep the cell coverage area of operator's planning, must keep load to be lower than specific level.This means that load estimation also is important for coverage.Particularly, it seems that in the fast dispatch of the uplink traffic that strengthens, load estimation is important in RBS from the coverage viewpoint.

And jam control function and permission control also benefit from the precise information about the instantaneous noise rising of each sub-district of its control in the radio network controller (RNC) of a plurality of RBS of control.The RNC function is slower than the bandwidth for the uplink scheduling that strengthens described above widely by its bandwidth that influences estate performance, yet, more than also be effective for the permission controlled function of RNC to a certain extent for the influence of the sub-district stability of the up link discussion that strengthens.

Permission control guarantees that the number of the user in the sub-district can not become aspect hardware resource and aspect load greater than the number that can be controlled.Too high load at first shows as too poor service quality, is controlled the fact of the increase that reaches sir target by the external power control loop.In principle, this feedback loop also may be introduced power and emerge in large numbers, as described above.

The permission controlled function can be avoided above influence by adjustment user's number with for the type of the corresponding traffic carrying capacity that each sub-district allowed that RNC controlled.The input that reaches the particular importance of this target is the accurate estimation that the noise of sub-district rises.

Can signal to RNC though the noise of estimating in RBS rises, all manufacturers may not support this signaling, or enough accurate load estimation may be provided.Therefore, need to estimate that the noise in RNC rises.

When the scheduling of the uplink traffic that strengthens was implemented in RNC, other problem appearred.Because RNC can control about 1000 sub-districts and in the future may be more much more, also will multiply by the number of the sub-district of service about the quite appropriateness requirement of memory consumption and the processing power algorithm that is used for the noise rise estimation today.Particularly, the solution of wasted memory is difficult to implement in RNC.Last very important benefit is that disclosed algorithm is applicable to the ASIC embodiment in the disclosure of invention.

Summary of the invention

The common problem of prior art cdma communication net is that load estimation provides with a kind of precision, this feasible load control that is difficult to carry out carefulness.Particularly, determine that what noise aspect the uplink channel that strengthens rose being subjected to mainly is because the big uncertainty that the difficulty of estimating noise substrate or other amount relevant with power causes.And the high memory requirement during noise floor estimation may be the obstacle that adds.

That general purpose of the present invention provides is improved, be used for determining the amount relevant with power, for example method and apparatus of load estimation.Other purpose of the present invention provides the method and apparatus of determining the amount relevant with power more accurately.A further object of the present invention provides the method and apparatus that is used to improve the noise rise estimation.Another object of the present invention provides the method and apparatus low requirement, that be used for determining the amount relevant with power that has for memory.

The above object is to finish by the method and apparatus according to Patent right requirement of the present disclosure.Generally speaking, in first aspect, provided the method that is used for the noise rise estimation in wireless communication system, this method comprises: the sample of measuring the total wideband power that receives at least.From the sample of the measurement of the total wideband power that receives at least, estimate probability distribution for first quantity of power.Typically, this first quantity of power is the total wideband power itself that receives.Be used for the conditional probability distribution of calculating noise substrate tolerance for the probability distribution of first quantity of power.This calculating is recursively carried out.Last numerical value according to the conditional probability distribution calculating noise rise measure of measuring for noise floor.

In second aspect, provide the node of wireless communication system.Typically, this node is RNC.This node comprise the total wideband power that is used to be received at least the measurement sample device and be used for estimating device for the probability distribution of first quantity of power from the sample of measuring at least of the total wideband power that receives at least.This node also comprise with recursive fashion operation, be used for according to device for the conditional probability distribution of the described at least probability distribution calculating noise substrate tolerance of first quantity of power.This node also comprises the device that is used for according to the numerical value of the conditional probability distribution calculating noise rise measure of measuring for noise floor.

Even an advantage of the invention is and when having adjacent area interference, external interference source and rapid fluctuations power, still can provide accurate noise rising value.And the present invention has quite low computational complexity and memory requirement.Additional advantage is discussed in conjunction with detailed description.

Description of drawings

By with reference to the following explanation of making together with accompanying drawing, can understand the present invention and other purpose of the present invention and advantage best, wherein:

Fig. 1 illustrates the signal chains of the radio base station of carrying out load estimation;

Fig. 2 be illustrated in the sub-district noise rise and total bit rate between typical relation;

Fig. 3 is the schematic diagram of the signal power that occurs in typical mobile radio communication;

Fig. 4 is the time diagram of the total wideband power of reception;

Fig. 5 is the block diagram according to the embodiment of noise rise estimation device of the present invention;

Fig. 6 is the diagram according to the recursive algorithm that interdepends of the present invention;

Fig. 7 is the figure that illustrates according to the performance of trace simulation of the present invention;

Fig. 8 is the figure that the performance of tracking when the background level flip-flop is shown;

Fig. 9 is the figure of probability density function that the minimum value of the quantity of power that measures from total received power is shown;

Figure 10 is the block diagram according to the major part of the embodiment of system of the present invention; And

Figure 11 is the flow chart according to the key step of the embodiment of method of the present invention.

Embodiment

In whole disclosures, the bold-type letter in formula is meant vector or matrix amount.

The complement of different distribution functions has been discussed in this disclosure.Be defined as follows.The complement of cumulative distribution function F is defined as 1 and deducts cumulative distribution function F.At for example accumulated error distribution function

Situation under, the complement of accumulated error distribution function becomes $1-F_{\mathrm{\Δx}(t\′|t\′)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\′|t\′)).$

This detailed description is by being introduced into about more deep a little discussion how to carry out the problem that load estimation and prior art solution run into, so that disclose its order of severity.This finishes with reference to typical WCDMA system, but described thought is not limited to WCDMA.But these thoughts can be applicable to the cdma system of the cellular system of many types, particularly all kinds.

Reference and measurement point

The type signal chain of RBS is shown in Fig. 1.At first regulate chain 2 from the broadband signal that antenna 1 receives by the analog signal that comprises cable, filter etc.Variation between the parts and temperature drift make when this signal enters receiver 3, the zoom factor of these parts of system the definite of 2-3dB that have an appointment.This further discusses below.In receiver 3, a plurality of operations take place.For load estimation, the total broadband power that receives of hypothesis is represented with 5 on Fig. 1 at certain one-level place usually, and is measured.And, suppose that in the present embodiment code power measures, i.e. the power of each channels/users of each of sub-district, 6 places can obtain in level.The reference point of the amount that is used to estimate is represented as 4.The amount of Gu Jiing is effectively and point that measure thereon, on the chain schematically is illustrated in Fig. 1 thereon.

Be difficult to estimate thermal noise floor power, several reasons is arranged.A reason is a thermal noise floor power as noted above, and the power of other reception, is subjected to the uncertain influence of analog receiver front center part.The signal reference point is at the antenna connector place according to definition.Yet, in digital receiver, after analog signal is regulated chain, obtain measuring.

Analog signal is regulated electronic chain is introduced 2-3dB really between RBS (criticizing) zoom factor error, and this is to be difficult to compensation.Measured by RTWP (total wideband power of reception) thereby may differ 2-3dB with the thermal noise power substrate of hypothesis divided by the default value of thermal noise power substrate.Influence is the also wrong 2-3dB of noise rise estimation.Consider the noise rise interval that allows in WCDMA system 0-7dB typically, the error of 2-3dB is unacceptable.

Fortunately, all power that form total received power are subjected to the same influence of zoom factor error γ (t), like this as calculating noise rising ratio N _R(t) time, the zoom factor error is cancelled as following formula:

$N_R\left(t\right)=N_R^{\mathrm{DigitalRe\; ceiver}}\left(t\right)=\frac{P^{\mathrm{Total},\mathrm{Digital\; Re\; ceiver}}\left(t\right)}{P_N^{\mathrm{Digital\; Re\; ceiver}}}=\frac{\mathrm{\γ}\left(t\right)P^{\mathrm{Total},\mathrm{Antenna}}\left(t\right)}{\mathrm{\γ}\left(t\right)P_N^{\mathrm{Antenna}}}=---\left(1\right)$

$=\frac{P^{\mathrm{Total},\mathrm{Antenna}}\left(t\right)}{P_N^{\mathrm{Antenna}}}=N_R^{\mathrm{Antenna}}\left(t\right)---\left(1\right)$

Wherein

With

Be respectively at digital receiver 3 (Fig. 1) and the noise rising ratio locating to measure at antenna 1 (Fig. 1), P ^{Total, DigitalReceiver}(t) and P ^{Total, Antenna}(t) be respectively at digital receiver 3 with at the total received power at antenna 1 place, and

With

Be respectively at digital receiver 3 and the thermal noise level measured at antenna 1 place.Yet, (1) formula that should be pointed out that need be in digital receiver noise floor

Measurement.This is a difficulty that is solved by the present invention.

Power measurement

In detailed description, used following general mark:

The measurement of total broadband power that receives is carried out in receiver.This measurement P ^Total(t) expression, wherein t represents discrete time.Measuring speed is T ^-1Hz.

Noise rises

As what indicate in background one joint, the result who introduces additional channel becomes the increase of gross power.Fig. 2 is the figure that these conditions are shown.Noise rising N _RBe the tolerance of load, it is defined in gross power and the thermal noise level P that measure, that be also referred to as noise floor at the antenna connector place _NBetween ratio.Surpassing noise rising threshold value N _R ^ThrThe time, situation becomes instability.At total bit rate and noise rising N _RBetween concern that 100 is known from the design of control ring, in case and instantaneous noise rising N _R ^ThrBe determined, just can carry out the scheduling of additional channel.Pole capacity C _PoleExpression is in the maximum bit rate capacity of bit per second.At threshold value N _R ^ThrWith by thermal noise level P _NTypical difference DELTA N between the level of regulation is 7dB.Yet, noise floor or thermal noise level P _NBe be not easy available.For example,, arrive 2-3dB greatly, thereby the major part of available surplus is subjected to the uncertainty influence of this introducing because the zoom factor uncertainty in the receiver is as discussed above.

The observability of noise floor

Occur being difficult to estimate a reason of thermal noise floor power now, even owing to make all measurements in digital receiver, noise floor can not directly be measured, and is not at least in single RBS.Explanation is adjacent area interference and also influences receiver from the interference of external source, and any mean value in such source can not be separated with noise floor.Power measurement on the cell channel of oneself can be carried out in some cases, increases the complexity of system.Yet such measurement does not solve whole issue, though what can improve this situation for they.

Fig. 3 illustrates the contribution for power measurement in conjunction with RBS 20.RBS 20 is associated with sub-district 30.In sub-district 30, have a large amount of portable terminals 25, it is communicated by letter with RBS 20 by different links, each with

Contribute to total received power.Sub-district 30 has in the intrasystem a plurality of adjacent sub-districts 31 of same WCDMA, and each sub-district is associated with RBS 21.Adjacent sub-district also comprises portable terminal 26.Portable terminal 26 emission radio-frequency powers, and the summation P of all such contributions ^NExpression.Other network-external radiation source can also be arranged, such as, for example radar station 41.Contribution P from such external source ^EExpression.At last, produce P from receiver itself _N.RBS 20,21 typically is connected to RNC 172.

From above can know see P ^N(t) and P _NBe immeasurablel, therefore need to be estimated in some way or be eliminated.Measurement iff total wideband power is available, then situation even become worse.Total wideband power is measured

Can be represented as:

$P_{\mathrm{Measurement}}^{\mathrm{Total}}\left(t\right)=\underset{i=1}{\overset{n}{\text{\Σ}}}{P}_i^{\mathrm{Code}}\left(t\right)+P^{E+N}\left(t\right)+P_N\left(t\right)+e^{\mathrm{Total}}\left(t\right),---\left(2\right)$

Wherein

P ^E+N＝P ^E+P ^N， (3)

And e wherein ^Total(t) give the measurement noise modeling.

Can on mathematics, prove P ^E+N(t) and P _NLinear estimation be not observable problem.Has only P ^E+N+ P _NAnd be observable from available measurement.This also is so under the situation of run time version power measurement.Problem is, do not have traditional technology can be used for noise floor with separate from adjacent area interference and interference source rises in the band of cellular system outside power average value.And, be available if having only the measurement of the broadband power of total reception, then each code power contribution also is undistinguishable with other contribution.

Noise floor estimation

Another reason of noise rise estimation difficulty is the amount that thermal noise floor is not always sought.Such situation is arranged, and interference influences the receiver of RBS significantly in the wherein constant band.These constant interference sources do not influence stability discussed above, but they are as the noise temperature appearance that increases, that is, and and the thermal noise floor of increase.

In the prior art, alternative be in this field, use each RBS thermal noise floor the height cost with each determine so that reach sufficiently high load estimation performance.The foundation of the default value of thermal noise power substrate, as what in digital receiver, see, the reference measure that need on a large amount of RBS, carry out in factory or ground on the scene.Two alternatives are expensive, and just need repeat as long as hardware changes.

The above solution of dealing with problems needs each RBS of each ground calibration.Yet this is very expensive, is extremely not have attraction therefore.And, in analog signal adjusting electronic circuit, will keep the perhaps temperature drift error of 0.7-1.0dB.

Another method provides the estimation of thermal noise power substrate.Be used to estimate that a principle of thermal noise power substrate is to estimate its minimum value as the quantity of power measurement that comprises thermal noise floor or that estimate.If without any the code power measurement is available, the broadband power that always receives typically of the quantity of power of being discussed then.Thereby method is that calculating noise rises being divided by as thermal noise floor power minimum value, that determine of the broadband power of instantaneous total reception and the broadband power that is estimated as total reception.

This schematically is shown in Fig. 4.The instantaneous value 102 of the total wideband power that receives is shown as the function of time here.This numerical value depends on instantaneous load and fluctuates greatly.As everyone knows, the thermal noise floor contribution always exists, and therefore can draw a conclusion, if measuring uncertainty is left in the basket, then the noise floor contribution must be equal to or less than the minimum value 104 of the broadband power of the total reception that receives in the regular hour section.The rational possibility if the contribution of all code power, neighbor cell contribution and other external contribution have equalled zero under certain chance, then minimum value 104 is good estimations of " really " noise floor 106.Yet, under all situations, can affirm that minimum value 104 constitutes the upper limit of unknown noise floor.

In order to improve the estimation of noise floor, can be applied to the measurement sequence to the recurrence estimation filter, the estimation of total wideband power of reception and its variance are provided.The thermal noise power substrate can be estimated by soft algorithm then.

Use has a plurality of features with the principle of being divided by of determined thermal noise floor power, and some feature wherein may be imperfect at least in some applications.Estimation principle determines that the special value of thermal noise power substrate is as output variable.This neither best neither be necessary.Really the output variable that needs is that noise rises, and as cited below, this amount can directly be estimated.And the valuation principle does not provide any measurement of precision of the thermal noise power substrate of estimation, does not provide noise to rise yet.This is owing to only estimated the result of a numerical value of thermal noise power substrate.

And above estimation principle does not have the probability distribution available existing information of consideration about for example real thermal noise floor power in the RBS set.This has further result.The estimation of the thermal noise power substrate that obtains by above thought always is biased and is higher than real numerical value.This is owing to thermal noise floor power, adjacent sub-district WCDMA power and non-WCDMA are with the always the same with the thermal noise power substrate at least with value big of interior interference power.Therefore, when estimating minimum value on the time interval of determining, always obtain numerical value greater than real thermal noise power.Consequently the noise rising is low estimation, that is, the load of sub-district is to hang down to estimate.The result can be too aggressive scheduling, and this causes for example sub-district unsteadiness.

As mentioned above, the present invention allows the conditional probability distribution that the direct estimation noise rises.This obtains as follows.The conditional probability distribution of the recursive algorithm estimating noise substrate of Ti Chuing in the present invention.And the probability distribution of the total wideband power of reception is available in the processing that further describes from below.By the formula of consideration for the merchant's of two stochastic variables distribution, the conditional probability distribution that can easily rise from these two probability distribution calculating noises.

Thus, additional important benefit of the present invention is the estimation of the One-dimensional probability function of noise rising, is not only single value, or determines the probability density function of noise floor at least.Estimate that the important benefit of probability distribution is to calculate the possibility of the variance of estimating (standard deviation) completely.Thus, the quality of estimation procedure will be assessed automatically.This uncertainty of picture is measured, and when uplink channel that scheduling in afterwards the step for example strengthens, may be useful especially.

The numerical value of noise floor can be estimated from the conditional probability distribution of noise floor really in the present invention.The noise basis floors of this estimation can be utilized to then by the valuation of the total wideband power of the current reception calculating noise rise measure value divided by the noise basis floors of estimating in a conventional manner.Yet such embodiment does not have above-mentioned advantage.

The embodiment that estimating noise rises is schematically illustrated with the block diagram form in Fig. 5.This embodiment relates to the load estimation field in the CDMA cellular telephone system.The disclosure of preferred embodiment is to write for the load estimation function about the up link of the enhancing in the cellular system of WCDMA type (E-UL).Yet, should be pointed out that the situation for other cellular system of CDMA type should be that similarly like this, the most of parts that go through also are effective for these systems.

Should be pointed out that in the following description probability distribution is by digital system, typically by discretely turning to histogram and processed distributing.

Noise rise estimation device 50 comprises three main pieces 60,70,80.In first power estimation block 60, the Kalman filter apparatus receives input 61, i.e. the total wideband power RTWP of the reception of measurement in the present embodiment.The mathematical details of preferred embodiment is revealed in appendix A.Output 69 from power estimation block 60 is the valuation and the corresponding variance of quantity of power, the i.e. valuation of the total wideband power RTWP of reception in the present embodiment and corresponding variance.Because output is from the Kalman filter apparatus, these parameters are unique parameters that need definition by the Gaussian Profile of the estimation of filter generation.Therefore, provide the total probability distributed intelligence that enough information defines the RTWP valuation.The filter details discusses in more detail below.

In more advanced system, power estimation block 60 can be according to other power parameter 62, for example the ratio (C/I) of the code power of the measurement of different channels i and interference, be used for channel i the β factor, be used for the code number of channel i and, carry out its estimation by the corresponding code power of fast power control ring order and the ratio of interference.In this case, the output 69 from power estimation block 60 can be the valuation and the corresponding variance of the relevant amount of another power.The valuation of quantity of power for example can be in neighbor cell WCDMA interference power, the band non-WCDMA interference power and thermal noise floor power and value.

Estimate in the piece 70 second conditional probability distribution,, and provide the output 79 that comprises the parameter that is associated with noise floor power based on valuation of the statistical device received power of Bayesian amount and corresponding standard deviation 69 conduct inputs.This can be the parameter of probability distribution of the estimation of the single numerical value of noise floor power or noise floor power.The histogrammic existing known parameters of the probability-distribution function of expression noise floor is stored in storage device 71, it provides the information 72 of probability distribution of the existing expection of relevant noise floor power to estimate piece 70 to conditional probability distribution, so that reach optimum evaluation.

Noise power substrate subsequently estimates that the effect of processing block is favourable, but beyonds one's depth.Provide highly technical explanation below for interested readers ' reading.

Should be pointed out that when the long-term average load of system increased, adjacent usually area interference increased.The result is that the possibility of the low numerical value of the gross power estimated increases with adjacent area interference and reduces.The calculating of the probability distribution of the minimum value of the gross power by applying estimation and move soft noise power substrate algorithm for estimating from the part of the existing probability distribution of above removal thermal noise power substrate.This existing center of gravity that distributes is shifted to lower numerical value, reduces the optimum evaluation of thermal noise power substrate thus.The amount of removing is determined by the probability distribution of the gross power that is in the estimation in the sliding window predetermined, sampling sparsely.The gross power probability distribution that has bigger variance then will deduct the bigger part of existing probability distribution significantly compared with the probability distribution with identical average and less variance.Reason is that the probability-distribution function with big variance further extends in the zone of the non-zero support that has probability distribution now.

Be used to estimate that the possible flat-footed method of minimum value is the valuation of calculating on preset time interval--so-called sliding window--.Detailed mathematical description based on the estimation of the conditional probability distribution of such sliding window provides in appendix B.

In the 3rd, noise rise estimation piece 80, the probability distribution 79 of the estimation of noise floor and RTWP valuation and corresponding standard deviation 68 are received as input, and the output 81 that comprises the noise rising value tentatively is provided.In the present embodiment, preferred noise rise measure is defined according to following formula:

$\mathrm{RoT}\left(t\right)=\frac{P^{\mathrm{Total}}\left(t\right)}{P_N},---\left(4\right)$

P wherein ^Total(t) be the total wideband power that receives, yet, also can utilize other noise rise measure.Further mention as above, what actual noise rose determines, the merchant's of the probability-distribution function that preferably can be by determining RTWP and the probability-distribution function of noise floor probability-distribution function is performed.

Piece 60,70 and 80 preferably can be integrated in the processor.Yet, also might use any device that includes, but not limited to different distributed solutions, can be looked at as the distributed treatment apparatus comprising the processor device of piece 60,70 and 80.

The estimation of the conditional probability distribution of the thermal noise floor that provides in appendix B is based on sliding window.These algorithms need be used to manage the parameter of sliding window size, because window size influences computational complexity.More importantly, this algorithm need be stored two matrix variables, takies the memory more than the 0.4-0.8Mbyte jointly.Particularly, for each power sample that is stored in the sliding window, on grid, need to calculate a probability-distribution function and a cumulative distribution function.Typically, grid on scope from-120dBm to-70dBm with the step-length of 0.1dB by discretization, for each power sample in sliding window, cause 1000 variablees.For 100 power sample in sliding window, the result needs storage 400000-800000 byte, depends on that being to use 4 bytes still is 8 byte variablees.

RNC can control about 1000 sub-districts in today, also will control more sub-districts in the future.Therefore, in RNC, may need 1000 examples (one of each sub-district).Memory consumption reaches the dynamic memory of 1GB, and this forbids.Conclusion is that the memory consumption of such algorithm that is used for soft noise floor estimation is unacceptable, at least for the RNC embodiment.Yet, may be to be feasible for the RBS embodiment that wherein only needs 4 examples of parallel running (4 diversity branches) for the memory requirement of sliding window method.

Should also be noted that because the renewal of noise floor only needs per minute to take place several times mean that the noise floor renewal for different districts can be scheduled in the different time intervals, the actual calculation complexity is unchallenged.

Second problem also relates to the use of the sliding window that is used for the minimum value estimation indirectly.Problem is owing to the power sample that enters sliding window, has a fractional value still remains on wherein at whole window duration.During this period, little numerical value nature is preponderated in minimum value is estimated.Therefore, under the situation that noise floor begins to increase, this did not reflect before the power sample with fractional value shifts out sliding window at last.

In order to solve above problem, the valuation of memory concerns, and feasible load particularly can be used to permit the purpose of control in RNC, and the present invention replaces the recursive algorithm that is used for soft noise floor estimation.

For first main thought of finding out suitable recursive algorithm is in the probability distribution of calculating minimum power, that is, introduce approximate during noise floor estimation.

All mark representations of using in the explanation are in this section at length explained in appendix B.The reason of this method is must read appendix B for the details of understanding this section.Roughly, t express time, x are represented (discretization) power, and f represents probability density function, and F represents cumulative distribution function.

First step that obtains recurrence formula is by considering that following situation removes the instantaneous influence of sliding window:

T _Lag→∞， (5)

That is, wherein the width of sliding window becomes infinity.

Then, the crucial formula (B12) of appendix B is transformed into:

$f_{\min {\left\{x_{P^{\mathrm{Total}}}^0(t\′)\right\}}_{t\′\≤t}|Y^t}\left(x\right)=\underset{t\′\≤t}{\mathrm{\Σ}}{f}_{\mathrm{\Δx}(t\′|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\′|t))\underset{\underset{q\≠t\′}{q\≤t}}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|t))).---\left(6\right)$

For following discussion, update time, t was by discretization,, introduced subscript N, so that provide that is

$f_{\min }(t_N,x)\≡f_{\min {\left\{x_{P^{\mathrm{Total}}}^0(t\′)\right\}}_{t\′\≤t_N}|Y^{t_N}}\left(x\right)$

$=\underset{t\′\≤t_N}{\mathrm{\Σ}}{f}_{\mathrm{\Δx}(t\′|t_N)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\′|t_N))\underset{\underset{q\≠t\′}{q\≤t_N}}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|t_N)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|t_N))).---\left(7\right)$

T wherein _NIt is the discretization time of upgrading.

Be introduced into first approximate according to following formula by using the filter valuation

Replace level and smooth valuation And obtain:

Suppose 1: ${\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\′|t_N)\≈{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\′|t\′)$

This hypothesis means that level and smooth gain is assumed to be little.In fact, approximate meaning, accept bad a little performance, so that the simplification that obtains calculating.Approximate expression 1 is reduced to formula (7):

$f_{\min }(t_N,x)\≈\underset{t\′\≤t_N}{\mathrm{\Σ}}{f}_{\mathrm{\Δx}(t\′|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\′|t))\underset{\underset{q\≠t\′}{q\≤t_N}}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|q)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|q))).---\left(8\right)$

Next procedure comprises the recurrence renewal of formulate complete long-pending (completed product).Complete long-pending Γ (t _N, x) be defined as:

$\mathrm{\Γ}(t_N,x)=\underset{\underset{}{q\≤t_N}}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|q)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|q))).---\left(9\right)$

Then, can represent complete amassing according to the following formula recurrence formula:

$\mathrm{\Γ}(t_{N+1},x)=\underset{q\≤t_{N+1}}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|q)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|q)))$

$=(1-F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))\underset{q\≤t_N}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|q)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|q))).---\left(10\right)$

$=(1-F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))\mathrm{\Γ}(t_N,x)$

This is first result, wherein should be pointed out that to calculate current complete long-pending Γ (t _N+1, x), that is, the product of the complement of the error profile of the accumulation of first quantity of power can be calculated as the complete long-pending Γ (t of former calculating _N, product x), that is, for the probability distribution of the accumulation of the complement of the error profile of the accumulation of first quantity of power and first quantity of power, according to the factor I of new complement $1-F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))$ The product of former calculating.

Next procedure is that the recurrence that obtains the probability density function of minimum power itself is upgraded, and, recursively writes out f that is _Min(t _N, x).This can draw as follows, from formula (8).

$f_{\min }(t_{N+1},x)\≈\underset{t\′\≤t_{N+1}}{\mathrm{\Σ}}{f}_{\mathrm{\Δx}(t\′|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\′|t))\underset{\underset{q\≠t\′}{q\≤t_{N+1}}}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|q)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|q)))$

$=f_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t_{N+1}|t_{N+1}))\underset{\underset{q\≠t_{N+1}}{q\≤t_{N+1}}}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|q)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|q)))$

$+\underset{t\′\≤t_N}{\mathrm{\Σ}}{f}_{\mathrm{\Δx}(t\′|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\′|t))\underset{\underset{q\≠t\′}{q\≤t_{N+1}}}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|q)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|q)))$

$=f_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t_{N+1}|t_{N+1}))\underset{q\≤t_N}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|q)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|q)))$

$+\underset{t\′\≤t_N}{\mathrm{\Σ}}{f}_{\mathrm{\Δx}(t\′|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\′|t\′))(1-F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))$

$\×\underset{\underset{q\≠r\′}{q\≤t_N}}{\mathrm{\Π}}(1-F_{\mathrm{\Δx}\left(q\right|q)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(q\right|q)))$

$=F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))\mathrm{\Γ}(t_N,x)$

$+(1-F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))f_{\min }(t_N,x).---\left(11\right)$

Here, can see noise floor tolerance f _Min(t _N+1, the calculating of the conditional probability distribution of renewal x) can be performed as two and value.First $f_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))\mathrm{\Γ}(t_N,x)$ Be the product Γ (t of former calculating of complement of error profile of the accumulation of first quantity of power _N, x) and factor $f_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))$ Product.This factor is as seeing based on the new probability distribution of first quantity of power.Second $(1-F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))f_{\min }(t_N,x)$ Be in complete long-pending recursive calculation, be used, noise floor tolerance before The conditions of calculation probability distribution f _Min(t _N, x) and factor I $1-F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))$ Product.

As conclusion, can see that the recursive calculation of the conditional probability distribution of noise floor tolerance is based on the product of former calculating of complement of error profile of accumulation of former calculating of former The conditions of calculation probability distribution, first quantity of power of noise floor tolerance and the new probability distribution of first quantity of power.The product of the complement of the error profile of the accumulation of first quantity of power also be according to the sum of products of the former calculating of the complement of the error profile of the accumulation of first quantity of power as the factor of the complement of the probability distribution of the new accumulation of first quantity of power by recursive calculation.In other words recursive calculation is the recursive calculation of the coupling of two amounts, that is, and and the product of the complement of the error profile of the conditional probability distribution of noise floor tolerance itself and the accumulation of first quantity of power.These are to be updated to the next necessary stored main entity that upgrades from one.Described main entity with the employed identical power grid by the sliding window algorithm on by discretization (seeing appendix B), yet the time scale of sliding window is removed.Compare with soft noise floor algorithm, can realize that memory requirement reduces 100 times based on the sliding window technology.This makes it possible in the permission control algolithm, even uses the disclosed algorithm that is used for load estimation in RNC.

Recursive calculation can be shown in the flow chart of Fig. 6 figure.The error profile of the current calculating of 800 expressions, first quantity of power.801, calculate the error profile of the accumulation of first quantity of power.According to the error profile of accumulation, factor I 804 is input to the recursive calculation 802 of the product of complement together with the product 805 of the former calculating of complement.The product 805 of the former calculating of complement also is combined into first 808 with factor 809, is used for the recursive calculation 803 of the conditional probability distribution of noise floor tolerance.To this calculate in 803 second 807 comprise factor I 804 and noise floor tolerance before The conditions of calculation probability distribution 806.

The recursion method that proposes involves approximate at present.Yet as what see on Fig. 7, approximate influence is almost insignificant.This accompanying drawing is the comparison between the disclosed recursive algorithm 701 in sliding window embodiment 700 and presents.Just as can be seen, consistency is good.Changing only is all sides of about 0.05dB.The performance of the variation of disclosed algorithm is because the adjusting of best tracking performance.

In its citation form, recursion method has some shortcomings.The most tangible shortcoming is to forget any characteristic of information in the past never fully.Thereby algorithmic statement is to stable state, and the condition of any drift or change all will have the problem that influences noise floor estimation after a little while.Therefore, preferably include certain data and forget mechanism.

First simple method that data are forgotten is only to interrupt this algorithm, and makes algorithm start from initial value again.This enabled condition changes, but reduces performance during first period after the startup.More or less exquisiter method is to make new recurrence start a moment later after old recurrence stops.Under such situation, new recurrence is approaching well real noise basis floors before in fact it be used.Shortcoming is that two parallel recurrence have a little while and all work, and this makes embodiment complicated.

Data are forgotten also can pass through recurrence discrete time filtering technique, for example by means of standard recurrence single order discrete time filter, and is introduced into.The bandwidth of the algorithm that finally obtains is directly controlled by the filter constants of recursion filter.For each fixing power network lattice point, recurrence formula (11) has the form of the introducing that is applicable to that immediately data are forgotten, f _Min(t _N, x) see state and Γ (t as _N, x) as input.Use 0＜β＜1 as filter constants, cause recurrence:

$f_{\min }(t_{N+1},x)=\mathrm{\β}(1-F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))f_{\min }(t_N,x)$

$+(1-\mathrm{\β})f_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))\mathrm{\Γ}(t_N,x).---\left(12\right)$

Recurrence (10) can not be made into the linear recurrence filtered version, because it continues to keep.Yet, by taking the logarithm, obtain following recurrence:

$\ln \left(\mathrm{\Γ}(t_{N+1},x)\right)=\ln {(1-F}_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))+\ln \left(\mathrm{\Γ}(t_N,x)\right).---\left(13\right)$

Then can be data be forgotten and be incorporated into (13) by using filter constants α.The result is:

$\ln \left(\mathrm{\Γ}(t_{N+1},x)\right)=(1-\mathrm{\α})\ln {(1-F}_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))+\mathrm{\α}\ln \left(\mathrm{\Γ}(t_N,x)\right)---\left(14\right)$

After getting index, obtain how much following filtering recurrence:

$\mathrm{\Γ}(t_{N+1},x)={(1-F_{\mathrm{\Δx}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))}^{1-\mathrm{\α}}\mathrm{\Γ}{(t_N,x)}^{\mathrm{\α}}---\left(15\right)$

Recursion (12) and (15) constitute final result.From the output of the recursion of these two couplings with as existing information in (B13) of appendix B combined, and calculate and carry out therefrom.

(12) and the initiation of (15) reach by following formula is set:

$\mathrm{\&Gamma;}({\mathrm{<figure-callout\; id="0"\; label="t"\; filenames="A200680055926E00331.png"\; state="\{\{state\}\}">t</figure-callout>}}_0,x)=1(\&DoubleRightArrow;\mathrm{\&Gamma;}({\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A200680055926E00301.png,A200680055926E00302.png"\; state="\{\{state\}\}">t</figure-callout>}}_1,x)={1-F}_{\mathrm{\&Delta;x}\left({\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A200680055926E00301.png,A200680055926E00302.png"\; state="\{\{state\}\}">t</figure-callout>}}_1\right|t_1)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left({\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A200680055926E00301.png,A200680055926E00302.png"\; state="\{\{state\}\}">t</figure-callout>}}_1\right|t_1)))---\left(16\right)$

$f_{\min }({\mathrm{<figure-callout\; id="0"\; label="t"\; filenames="A200680055926E00331.png"\; state="\{\{state\}\}">t</figure-callout>}}_0,x)=0(\&DoubleRightArrow;f_{\min }({\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A200680055926E00301.png,A200680055926E00302.png"\; state="\{\{state\}\}">t</figure-callout>}}_1,x)=F_{\mathrm{\&Delta;x}\left({\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A200680055926E00301.png,A200680055926E00302.png"\; state="\{\{state\}\}">t</figure-callout>}}_1\right|t_1)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left({\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A200680055926E00301.png,A200680055926E00302.png"\; state="\{\{state\}\}">t</figure-callout>}}_1\right|t_1))),---\left(17\right)$

This is correct initial behavior.

Also having other method to introduce data forgets.A possible method is to use the propagation at random of the probability density function of (11).This will need the dynamic model hypothesis of the diffusion of probability density function then.Described method is quite complicated, does not make detailed process here.

Quote the recursive algorithm that is used for soft noise floor estimation and have several advantages.An advantage is, this algorithm only need be made an appointment with each sub-district 0.005M bytes of memory device, that is, compare with sliding window method about 1%.Recursive algorithm is compared with the sliding window algorithm, also reduces computational complexity.They have avoided controlling with restriction on the parameters the needs of computational complexity, also reduce the number of the parameter that is used to manage thus widely.They also allow to use α and β to regulate parameter and regulate by considering the standard engineering bandwidth.

The tracking characteristics of recursive algorithm can further be improved.Can introduce the particular procedure of some threshold parameter, so that on the dynamic range of non-constant width, obtain good tracking characteristics.

For first interpolation is described, should be pointed out that during iteration grid point places a lot of more than the broadband power that the numerical value of the probability density function of minimum power is measured become very little in the sub-district.It in addition can be zero in the resolution of computer arithmetic.As long as thermal noise floor does not change, this is acceptable.Yet under the situation that the thermal noise power substrate increases suddenly, the very little numerical value that is reduced to the probability density function of the broadband power that is lower than measurement after noise floor changes becomes at them and will need before 1 to approach the very long time to increase.As a result, under the situation that noise floor increases, follow-up control will be very poor.Actual change can spent the very long time before noticing fully thus.

In order to offset this undesired performance, quote the minimum License Value of the probability density function of minimum power.Any calculating of littler numerical value will be switched to minimum value.Typically, find that about 0.000001 numerical value is suitable.

Studied the tracking performance of the algorithm that proposes.As what can see on Fig. 8, this algorithm is successfully followed the tracks of the thermal noise power substrate and is surpassed 50dB.Power changes to be introduced when 1000s.

The probability density function of noise floor is shown in Fig. 9.If obtain less numerical value in calculating, all numerical value of probability density function is changed to 0.000001.Shown situation illustrates wherein probability density function 600 increases at pact-75dBm place, and it reduces at pact-110dBm place 601.Though probability density function is bigger at pact-110dBm place 601, surpass the conditional mean of probability distribution at the probability density function at pact-75dBm place 600, it equals optimum evaluation.This be because-75dBm, 600, be much bigger power, and because the peak value of pact-110dBm, the 601st, narrower.Should be pointed out that the peak value at pact-110dBm place 601 is that it disappears the most at last from the residue of beginning period.

Yet the result of above change is when the valuation of thermal noise power substrate is estimated, to introduce undesired skew.The origin of described skew is the artificial high numerical value of the probability density function of the common minimum power of introducing in most of grid points.These high numerical value cause the domination of high power grid point in conditional mean, make the fact of the noise power substrate that oneself shows as too high estimation.

Fortunately, this back problem can only be in the power network lattice point of minimal value level and is careful by removal from all calculating of conditional mean.In other words, for replacing, the grid point of estimating the thermal noise power substrate, be reduced to be lower than minimum value is set to equal zero.Should be pointed out that this also should be employed when when using the merchant to distribute to calculate soft noise rising valuation.

Arithmetic addition enables to follow the tracks of the input power greater than 50dB.This makes again might handle the recurrent RBS of configuration mistakenly in the WCDMA network effectively.The RBS of configuration mistakenly like this can see-120dBm and-man made noise's substrate between the 70dBm.And, can avoid needs for the safety net value, this safe net value is needed for various sliding window algorithms.These safe net value are introduced and are used for the further logic of the thermal noise floor of control estimation.

Should point out emphatically, the strong action need for effective load estimation is arranged in the permission controlled function of RNC.Because configuration error and front end zoom factor mistake in RBS are needing a large amount of artificial work for permitting control algolithm to regulate at present in field system.

Important advantage is that disclosed algorithm is applicable to the ASIC embodiment in the disclosure of invention.This is because algorithm as recursion filter work, does not need dynamic memory allocation.This is true to make the algorithm that is proposed be applicable to that afterwards RBS replaces the algorithm based on the RBS of sliding window of planning now in discharging, if this should be considered to cost economical.

In the above description, suppose that the power estimation relates to uplink communication.Power measurement is by the node in the radio access network under such situation, is typically carried out by radio base station or radio network controller.Figure 10 illustrates the major part according to the embodiment of system of the present invention, and wherein load estimation is carried out in RNC.Wireless communication system 170 comprises Universal Mobile Telecommunications System terrestrial radio Access Network (UTRAN) 171.RBS20 among portable terminal 25 and the UTRAN 171 carries out wireless link.RBS 20 is by radio network controller (RNC) 172 controls, and it is connected to mobile switching center/visitor location register (MSC/VLR) 174 and the service general packet radio support node (SGSN) 175 of serving of core net CN 173 again.

In the present embodiment, RBS 20 comprises power sensing device 51, and typically, antenna and front-end electronics are used to measure the total wideband power of instantaneous reception.Connect 53, so-called Iub interface is used in communicating by letter between RBS 20 and RNC 172.According to standard, the Iub interface permission transmits the sample of the measurement of the total wideband power that receives.Connecting 53 therefore is to be used for the device of data of sample of measurement that RNC 172 obtains representing the total wideband power of reception.Noise rise estimation device 50 is available in RNC172, is arranged for by connecting 53 samples that receive the measurement of the total wideband power received.

Figure 11 illustrates the flow chart according to the key step of the embodiment of method of the present invention.This program begins in step 200.In step 202, measure a plurality of samples of the total wideband power that receives at least.In step 210, estimate probability distribution for first quantity of power from the sample of the measurement of the total wideband power that receives at least.First quantity of power can be the total wideband power that receives.In step 214, according to the conditional probability distribution of measuring for the probability distribution calculating noise substrate of first quantity of power at least.This step is recursively carried out.At last, in step 218, basis is for the numerical value of the conditional probability distribution calculating noise rise measure of noise floor tolerance at least.This program finishes in step 299.

Embodiment described above is appreciated that several illustrative example of the present invention.It will be understood to those of skill in the art that for these embodiment can modify, make up and change, and do not deviate from scope of the present invention.Particularly, the different parts solutions in different embodiment under the possible technically occasion, can be combined into other configuration.Yet scope of the present invention is defined by the following claims.

Appendix A:

Be used for the Kalman filter that RTWP measures

The algorithm that proposes for the situation of wherein measuring total RTWP is a prediction-renewal filter, and wherein subscript is distinguished prediction steps and step of updating.

$K_{\mathrm{Update}}\left(t\right)=\frac{P_{\mathrm{Prediction}}^{\mathrm{Cov}}(t-T_{\min })}{P_{\mathrm{Prediction}}^{\mathrm{Cov}}(t-T_{\min })+r_{\mathrm{Measurement}}}---\left(A1\right)$

$P_{\mathrm{Update}}^{\mathrm{Total}}\left(t\right)=P_{\mathrm{Prediction}}^{\mathrm{Total}}(t-T_{\min })+K_{\mathrm{Update}}\left(t\right)\&times;(P_{\mathrm{Measurement}}^{\mathrm{Total}}\left(t\right)-P_{\mathrm{Pr\; ediction}}^{\mathrm{Total}}\left(t\right))---\left(A2\right)$

$P_{\mathrm{Update}}^{\mathrm{Cov}}\left(t\right)=P_{\mathrm{Pr\; ediction}}^{\mathrm{Cov}}(t-T_{\min })-\frac{P_{\mathrm{Pr\; ediction}}^{{\mathrm{Cov}}^2}(t-T_{\min })}{P_{\mathrm{pr\; ediction}}^{\mathrm{Cov}}(t-T_{\min })+r_{\mathrm{Measurement}}}---\left(A3\right)$

$P_{\mathrm{Pr\; ediction}}^{\mathrm{Total}}\left(t\right)=P_{\mathrm{Update}}^{\mathrm{Total}}\left(t\right)---\left(A4\right)$

$P_{\mathrm{Pr\; ediction}}^{\mathrm{Cov}}\left(t\right)=P_{\mathrm{Update}}^{\mathrm{Cov}}\left(t\right)+\frac{T_{\min }}{T_{\mathrm{Correlation}}}r---\left(A5\right)$

(A1)-(A5) by t is increased T _MinStep-length and repeat.

Carry out initialization at t=0 by following formula:

$P_{\mathrm{Pr\; ediction}}^{\mathrm{Total}}\left(0\right)={\mathrm{<figure-callout\; id="0"\; label="P"\; filenames="A200680055926E00331.png"\; state="\{\{state\}\}">P</figure-callout>}}_0^{\mathrm{Total}}---\left(A6\right)$

$P_{\mathrm{Pr\; ediction}}^{\mathrm{Cov}}\left(0\right)={\mathrm{<figure-callout\; id="0"\; label="P"\; filenames="A200680055926E00331.png"\; state="\{\{state\}\}">P</figure-callout>}}_0.---\left(A7\right)$

See as above, more new gain K _Update(t) be from model parameter r _MeasurementCovariance with the prediction that from former sampling example, obtains

Calculate.Then, by using prediction

With new measurement Calculate the total wideband power of upgrading with up-to-date measurement

Next procedure is covariance and the r from prediction _MeasurementCalculate the covariance of upgrading

In last iterative step, calculate

With

New numerical value, and add step-length to the time.T _MinThe expression sampling period.

Appendix B

The estimation of conditional probability distribution

Be noted that and estimate that minimum power is very natural.Yet, use the selection of minimum value to be actually specific (ad-hoc).Under common situation, the extremum of amount depends on the P of estimation in some way ^TotalAmount might be used as the basis that is used for other calculating.Yet, as simple embodiment, consideration amount here

Should be pointed out that and be about in question P ^TotalThe total wideband power that finger is received.

Representation, conditional probability and Baye rule

Be widely used in the Bayes rule of probability distribution and the definition of conditional mean below.Following definition and result can be for example at T.

, " Discrete Time StochasticSystems ", London, UK:Springer, 2002, find in 12-14 page or leaf or any other textbook about estimation.

Probability distribution: consider two incident A and B, have probability distribution f respectively _A(x) and f _B(y).Then, the joint probability distribution of A and B is represented as f _{A, B}(x, y).

The incident that should be pointed out that is represented by subscript with regulating, and independently variable appears in the bracket.This representation only is used when the probability distribution of probability of use distribution and accumulation.When the State Estimation of mentioning Kalman filter for example and covariance, in bracket, also occur regulating.

Conditional probability distribution: conditional probability distribution f _A|B(x) and f _B|A(y) define by following formula:

f _A，B(x，y)＝f _A|B(x)f _B(y)＝f _B|A(y)f _A(x). (B1)

Should be pointed out that as the result who is used for the representation of probability distribution, regulate also being represented as subscript.

The solution of above formula causes famous Bayes rule now:

$f_{A|B}\left(x\right)=\frac{f_{B|A}\left(y\right)f_A\left(x\right)}{f_B\left(y\right)}.---\left(B2\right)$

Should be pointed out that and to understand above rule best by using the circle diagram that intersects.The formal proof that obtains the result of probability distribution for example can be used the infinitesimal limited version of the motivation that is used for the probability situation.

The conditional probability of minimum value--model and universal expression formula

In this section, some generic features of derivation minimum value estimator for this reason, is introduced following representation.Kalman filter or P ^TotalThe valuation of the Kalman smoother of (t ') is expressed from the next:

${\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|Y^t)\&equiv;{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|{\left\{y\left(s\right)\right\}}_{s\&Element;[-\&infin;,t]})$

$={\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|{\left\{y\left(s\right)\right\}}_{s\&Element;[t-T_{\mathrm{Lag}},t]},{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t-T_{\mathrm{Lag}}|Y^{t-T_{\mathrm{Lag}}})).---\left(B3\right)$

Here, t ' is illustrated in Certain interior time.Condition distributes, and under moderate condition, all is that Gauss enough adds up, and, only needs second-order characteristics in order to describe conditional probability distribution that is.This is reflected in the adjusting in the last expression formula of [A3].Be distributed as with postcondition:

$f_{{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;)|Y^t}\left(x\right)\&Element;N({\hat{x}}_{P^{\mathrm{Tatal}}}^{\mathrm{Kalman}}(t\&prime;|t),{\left({\mathrm{\&sigma;}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t)\right)}^2),---\left(B4\right)$

Wherein Kalman represents that valuation is by the Kalman filter, if or t '＜t, calculate by the Kalman smoother.Amount With

Represent power valuation and corresponding covariance respectively, that is, and to the input of estimator.Should be pointed out that [B4] supposes at time t-T _LagCorresponding valuation be used as the initial value of Kalman filter.

Then, further the condition of the minimum value of developed horse power valuation distributes.For this reason, at the real power of representative $x_{P^{\mathrm{Total}}}^O\left(t^{\&prime;}\right)=P^{O,\mathrm{Total}}\left(t^{\&prime;}\right)$ With represent valuation ${\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t)={\hat{P}}^{\mathrm{Total}}(t\&prime;|t)$ Between relation, suppose following model:

$x_{P^{\mathrm{Total}}}^0(t\&prime;)={\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t)+\mathrm{\&Delta;}{x}_{P^{\mathrm{Total}}}(t\&prime;|t)---\left(B5\right)$

$x_{P^{\mathrm{Total}}}^0(t\&prime;)=N({\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t),{\left({\mathrm{\&sigma;}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t)\right)}^2).---\left(B6\right)$

This is consistent with above discussion for enough statistics.For

The representation of distribution therefore be reduced to:

f _Δx(x). (B7)

Should be pointed out that this distribution needn't be assumed to be Gaussian (though this is the hypothesis of making most).

Then, by using the data y (t) that obtains from the time interval [∞, t] to estimate $x_{P^{\mathrm{Total}}}^0(t\&prime;)={\mathrm{<figure-callout\; id="0"\; label="P"\; filenames="A200680055926E00331.png"\; state="\{\{state\}\}">P</figure-callout>}}^{0,\mathrm{Total}}(t\&prime;),$

The conditional probability distribution of minimum value.

Fig. 4 provides the broadband power P that total reception is shown ^Total(t) time changes 102 figure.During some time interval, total broadband power that receives presents high numerical value.Yet under some chance, total broadband power that receives becomes little, shows for the many common contribution of measuring power not exist.

As what it will be appreciated that below, need smoother valuation conduct in theory to being used in the time interval

The input of the conditional probability estimation algorithm of the minimum power of last work.For formal maintenance optimality under development, the smoother valuation also should be by using

All interior data are calculated.Yet in the embodiment of reality, typically only the of short duration snapshot to data calculates near selected smoothingtime point by using in these smoother valuations.From

Several so level and smooth valuations be combined then, be used for estimating conditional probability distribution.In being about to discussion, although the time interval

Be maintained in all amounts, so that exploitation is not too complicated.By replacing the smoother valuation, can reach further simplification with the valuation of Kalman filter.Simulation shows that this can finish with very little performance loss.

The condition of minimum value distributes and can be written as following [consulting (B5)] now:

$f_{\min {\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag},t}]}|Y^t,\min x_{P^{\mathrm{Total}}}^0(t-T_{\mathrm{Lag}})}\left(x\right),---\left(B8\right)$

Wherein the last scale of [B8] shows the initial information of minimum value.Following, Bayes rule and be widely used for the definition of the conditional mean of probability distribution.

Then, by using, the definition of Bayes rule and conditional probability is applied to (B8) to give a definition:

$A:=\min {\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}$

$B:={\min x}_{P^{\mathrm{Total}}(t-T_{\mathrm{Lag}})}^0$

C：＝Y ^t

Definition by using Bayes rule, conditional probability distribution and f as a result _{B, C|A}(x, y)=f _{(B|A). (C|A)}(x y), keeps following equation chain (latter's result is easy to be verified by the drawing of three-circle diagram):

$f_{A|B,C}\left(x\right)=\frac{f_{B,C|A}(x,y)f_A\left(x\right)}{f_{B,C}(x,y)}=\frac{f_{\left(B\right|A),\left(C\right|A)}(x,y)f_A\left(x\right)}{f_{B,C}(x,y)}$

$=\frac{f_{\left(B\right|A)|\left(C\right|A)}\left(x\right)f_{C|A}\left(y\right)f_A\left(x\right)}{f_{B,C}(x,y)}=\frac{f_{B|A,C}\left(x\right)f_{C|A}\left(y\right)f_A\left(x\right)}{f_{B,C}(x,y)}$

$=\frac{f_{B|A,C}\left(x\right)f_{A|C}\left(x\right)f_C\left(y\right)}{f_{B,C}(x,y)}.---\left(B9\right)$

Last step is easy to be verified by the drawing three-circle diagram once more.Now, according to above definition, the factor I in the molecule (B9) is a priori, therefore regulates to disappear.The factor of molecule will further specify below, and the part that the last factor of molecule and denominator can be used as normaliztion constant is treated.A, the replacement following relation of proof then of returning of the definition of B and C:

$f_{\min {\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}|Y^t,\min x_{P^{\mathrm{Total}}}^0(t-T_{\mathrm{Lag}})}\left(x\right)$

$=\frac{1}{c}{f}_{\min {\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}|Y^t,}\left(x\right)f_{\min x_{P^{\mathrm{Total}}}^0(t-T_{\mathrm{Lag}})}\left(x\right).---\left(B10\right)$

The result of (B10) that need remember is that smoothing problasm is on the horizon.Therefore the pre-treatment step of more than treating based on Kalman filtering need comprise Kalman smoother step in form.In fact, however the Kalman filter is normally enough.

The last expansion of the conditional mean of minimum power

Starting point of this son joint is formula (B10), and its sets forth conditions pdf (probability-distribution function) is presented as preceding value (initial value) and the product that depends on the factor of measurement.Preceding value is provided by the user, and it should reflect previous about P _NUncertainty.No matter when should be pointed out that sliding window is moved when being calculated with new valuation, uses same preceding value once more.Therefore preceding value is not updated in the basic setup of estimator.

In order to set forth condition pdf completely, need some further processing of the factor I of (B10).(B7) error profile f _{Δ P}(x) and definition (B5) and (B6), will be the center for this reason.And, in following calculating, the distribution of F () expression accumulation, that is, and the integration of f.The probability of Pr (.) presentation of events.

Following equation is set up for the factor I of (B10) now:

$F_{\min {\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}|Y^t}\left(x\right)=\mathrm{Pr}(\min x{\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}\&le;x|Y^t)$

$=1-\mathrm{Pr}(\min {\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}>x|Y^t)$

$=1-\mathrm{Pr}(\&ForAll;t\&prime;,\mathrm{\&Delta;}{x}_{P^{\mathrm{Total}}}(t\&prime;|t)>x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t))$

$=1-\underset{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}{\mathrm{\&Pi;}}\mathrm{Pr}(\mathrm{\&Delta;}{x}_{P^{\mathrm{Total}}}(t\&prime;|t)>x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t))$

$=1-\underset{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}{\mathrm{\&Pi;}}(1-\mathrm{Pr}(\mathrm{\&Delta;}{x}_{P^{\mathrm{Total}}}(t\&prime;|t)\&le;x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t)))$

$=1-\underset{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}{\mathrm{\&Pi;}}(1-F_{\mathrm{\&Delta;x}(t\&prime;|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t))).---\left(B11\right)$

(B11) fourth class formula provides the hypothesis of enough statistic from the Kalman smoother, that is, draw from (B5) with (B6).Last equation draws from (B7).Obviously, the most natural hypothesis is to use for F _{Δ P (s)}Gaussian Profile.Yet, (B11) in fact also allow other distribution.

Final step in the derivation of the factor I of distribution function is differential (B11), obtains:

$F_{\min {\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}|Y^t}\left(x\right)=\frac{{\mathrm{dF}}_{\min {\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag}},]t}|Y^t}\left(x\right)}{\mathrm{dx}}$

$=\underset{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}{\mathrm{\&Sigma;}}{f}_{\mathrm{\&Delta;x}(t\&prime;|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t))\underset{q\&NotEqual;t\&prime;}{\underset{q\&Element;[t-T_{\mathrm{Lag}},t]}{\mathrm{\&Pi;}}}(1-F_{\mathrm{\&Delta;x}(t\&prime;|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|q)))---\left(B12\right)$

Combined with (B10), provide final result:

$f_{\min {\left\{x_{P^{\mathrm{Total}}}^0(t\&prime;)\right\}}_{t\&prime;\&Element;[t-T_{\mathrm{Lag},t}]}|Y^t,\min x_{P^{\mathrm{Total}}}^0(t-T_{\mathrm{Lag}})}\left(x\right)$

$=\frac{1}{c}\left(\underset{t\&prime;\&Element;[t-T_{\mathrm{Lag}},t]}{\mathrm{\&Sigma;}}{f}_{\mathrm{\&Delta;x}(t\&prime;|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t))\underset{q\&NotEqual;t\&prime;}{\underset{q\&Element;[t-T_{\mathrm{Lag}},t]}{\mathrm{\&Pi;}}}(1-F_{\mathrm{\&Delta;x}(t\&prime;|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|q)))\right)f_{\min x_{P^{\mathrm{Total}}}^0(t-T_{\mathrm{Lag}})}\left(x\right)---\left(B13\right)$

This result constitutes the output 79 that relates in conjunction with Fig. 5.Expression formula seems may be very complicated.Fortunately, it can be estimated straight from the shoulder, because it is the Gaussian Profile of one dimension Gaussian function and accumulation, is given:

$F_{\mathrm{\&Delta;x}(t\&prime;|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t))=\frac{1}{\sqrt{2\mathrm{\&pi;}}{\mathrm{\&sigma;}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t)}{e}^{-\frac{{(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t))}^2}{2{\left({\mathrm{\&sigma;}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t)\right)}^2}}---\left(B14\right)$

$F_{\mathrm{\&Delta;x}(t\&prime;|t)}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t))=\underset{-\&infin;}{\overset{x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t)}{\&Integral;}}{f}_{\mathrm{\&Delta;x}(t\&prime;|t)}\left(y\right)\mathrm{dy}$

$=\frac{1}{2}\mathrm{erfc}(-\frac{(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t))}{\sqrt{2}{\mathrm{\&sigma;}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}(t\&prime;|t)}).---\left(B15\right)$

Amount

With

Easily as from the output of Kalman smoother or better simply Kalman filter and available.

If provide the noise basis floors, then carry out mean value computation for the distribution of output as output.

Claims (21)

1. in wireless communication system, be used for the method for noise rise estimation, may further comprise the steps:

Measure the sample of the total wideband power that receives at least;

From the sample of the described at least measurement of the total wideband power that receives at least, estimate probability distribution for first quantity of power;

According to the conditional probability distribution of measuring for the described probability distribution calculating noise substrate of described first quantity of power at least;

The step of described calculating recursively is performed; And

Numerical value according to the described conditional probability distribution calculating noise rise measure of measuring for described noise floor.

2. according to the process of claim 1 wherein that the described recursive calculation of described conditional probability distribution of described noise floor tolerance is based on the sum of products of former calculating of the complement that the accumulated error of the former calculating of the former The conditions of calculation probability distribution of described noise floor tolerance, described first quantity of power distributes for the new probability distribution of described first quantity of power.

3. according to the method for claim 2, the described recursive calculation of the described conditional probability distribution of wherein said noise floor tolerance is based on the recursive calculation of product of the described calculating of the complement that the accumulated error of the former calculating of described first quantity of power distributes.

4. according to the method for claim 3, the step of the described conditional probability distribution of the described noise floor tolerance of wherein said recursive calculation may further comprise the steps again:

The product of the former calculating of the complement that the present product that calculates the complement that the described accumulated error of described first quantity of power distributes distributes as the described accumulated error of described first quantity of power and the product of the factor I of the new complement that distributes based on described cumulative probability for described first quantity of power; And

The described conditional probability distribution of calculating described noise floor tolerance as first with second and value, described first be the product that calculates before the complement that distributes of the described accumulated error of described first quantity of power described with based on product for the factor of the new probability distribution of described first quantity of power, described second is the product of The conditions of calculation probability distribution and described factor I before described noise floor tolerance described.

5. according to the method for claim 4, the step of the described conditional probability distribution of the described noise floor tolerance of wherein said recursive calculation is carried out according to following formula:

$f_{\min }(t_{N+1},x)=f_{\mathrm{\&Delta;x}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1}))\mathrm{\&Gamma;}(t_N,x)$

$+(1-F_{\mathrm{\&Delta;x}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))f_{\min }(t_N,x),$

$\mathrm{\&Gamma;}(t_{N+1},x)=(1-F_{\mathrm{\&Delta;x}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))\mathrm{\&Gamma;}(t_N,x),$

T wherein _NBe the Measuring Time of the sample N of the total wideband power that receives at least, x represents the power of discretization, f _Min(t _N, be x) at time t _NThe probability density function of the minimum value of the above first quantity of power, Γ (t _N, x) be the described product of the complement that distributes of the described accumulated error of described first quantity of power, $f_{\mathrm{\&Delta;x}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))$ Be at time t _N+1The error profile of the above first quantity of power, and $F_{\mathrm{\&Delta;x}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))$ Be at time t _N+1The accumulated error of the above first quantity of power distributes.

6. according to each method of claim 1 to 5, comprise and introduce the other step that data are forgotten mechanism.

7. according to the method for claim 6, wherein said introducing data are forgotten that the step of mechanism comprises and are restarted described noise rise estimation off and on.

8. according to the method for claim 6, wherein said introducing data forget that the step of mechanism comprises the described conditional probability density function of the tolerance of borne noise substrate at random.

9. according to the method for claim 4 and 6, wherein said data forget that mechanism implements with filter constants in the recursive calculation step.

10. according to the method for claim 5 and 9, wherein said data forget that mechanism is implemented as:

$\mathrm{\&Gamma;}(t_{N+1},x)={(1-F_{\mathrm{\&Delta;x}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))}^{1-\mathrm{\&alpha;}}\mathrm{\&Gamma;}{(t_N,x)}^{\mathrm{\&alpha;}},$

$f_{\min }(t_{N+1},x)=\mathrm{\&beta;}{(1-F_{\mathrm{\&Delta;x}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))}^{}{f}_{\min }(t_N,x)$

$+(1-\mathrm{\&beta;}){f_{\mathrm{\&Delta;x}\left(t_{N+1}\right|t_{N+1})}(x-{\hat{x}}_{P^{\mathrm{Total}}}^{\mathrm{Kalman}}\left(t_{N+1}\right|t_{N+1})))}^{\mathrm{}}\mathrm{\&Gamma;}(t_N,x),$

Wherein α and β are filter constants.

11., also comprise the other step of the minimum value of the described conditional probability distribution of quoting described noise floor tolerance according to each method of claim 1 to 10.

12. according to the method for claim 11, wherein said minimum value is 0.000001 the order of magnitude.

13. according to the method for claim 11 or 12, the power network lattice point of described conditional probability distribution that wherein has the described noise floor tolerance of described minimum value is removed from the step of described basis for the numerical value of the described conditional probability distribution calculating noise rise measure of noise floor tolerance.

14. according to each method of claim 1 to 13, the step of the numerical value of wherein said calculating noise rise measure is based on the valuation of noise floor, it is again based on the described conditional probability distribution of described noise floor tolerance.

15. according to each method of claim 1 to 13, the step of the numerical value of wherein said calculating noise rise measure is based on the conditional probability distribution of described noise rise measure, it is again based on the described conditional probability distribution of described noise floor tolerance.

16. according to each method of claim 1 to 15, wherein said first quantity of power is the total wideband power that receives.

17. the node of wireless communication system comprises:

The device of the measurement sample of the total wideband power that is used to be received at least;

Be used for estimating device, be connected to the described device of the measurement sample of the total wideband power that is used to be received at least for the probability distribution of first quantity of power from the reception sample of the described at least measurement of total wideband power at least;

Be used for according to device, be connected to the described device that is used to estimate for the probability distribution of first quantity of power for the conditional probability distribution of the described at least probability distribution calculating noise substrate tolerance of described first quantity of power;

The described device that is used for the conditional probability distribution of calculating noise substrate tolerance is arranged to recursively carry out described calculating; And

Be used for being connected to the described device that is used for the conditional probability distribution of calculating noise substrate tolerance according to the device that calculates the numerical value of described noise rise measure for the described conditional probability distribution of described noise floor tolerance.

18. according to the node of claim 17, the device of the sample of the wherein said measurement that is used to the total wideband power that obtains receiving comprises the device of data of sample that is used for receiving by communication interface the measurement of the total wideband power that representative receives at least.

19. according to the node of claim 17 or 18, wherein said node is a radio network controller.

20. according to each node of claim 17 to 19, wherein said node is the node of WCDMA system.

21. wireless communication system comprises:

Each at least one node according to claim 17 to 20.

Images